[All] ImageOcr (#662)

Lagrange.Core/Common/Entity/ImageOcrResultEntry.cs

@@ -0,0 +1,49 @@
+using System.Text.Json.Serialization;
+
+namespace Lagrange.Core.Common.Entity
+{
+    [Serializable]
+    public class ImageOcrResult
+    {
+        [JsonPropertyName("texts")] public List<TextDetection> Texts { get; set; }
+
+        [JsonPropertyName("language")] public string Language { get; set; }
+
+        public ImageOcrResult(List<TextDetection> texts, string language)
+        {
+            Texts = texts;
+            Language = language;
+        }
+    }
+
+    [Serializable]
+    public class TextDetection
+    {
+        [JsonPropertyName("text")] public string Text { get; set; }
+
+        [JsonPropertyName("confidence")] public int Confidence { get; set; }
+
+        [JsonPropertyName("coordinates")] public List<Coordinate> Coordinates { get; set; }
+
+        public TextDetection(string text, int confidence, List<Coordinate> coordinates)
+        {
+            Text = text;
+            Confidence = confidence;
+            Coordinates = coordinates;
+        }
+    }
+
+    [Serializable]
+    public class Coordinate
+    {
+        [JsonPropertyName("x")] public int X { get; set; }
+
+        [JsonPropertyName("y")] public int Y { get; set; }
+
+        public Coordinate(int x, int y)
+        {
+            X = x;
+            Y = y;
+        }
+    }
+}

Lagrange.Core/Common/Interface/Api/OperationExt.cs

@@ -261,4 +261,13 @@ public static Task<bool> GroupJoinEmojiChain(this BotContext bot, uint groupUin,
 
     public static Task<bool> FriendJoinEmojiChain(this BotContext bot, uint friendUin, uint emojiId, uint targetMessageSeq)
         => bot.ContextCollection.Business.OperationLogic.FriendJoinEmojiChain(friendUin, emojiId, targetMessageSeq);
+
+    public static Task<string> UploadImage(this BotContext bot, ImageEntity entity)
+        => bot.ContextCollection.Business.OperationLogic.UploadImage(entity);
+
+    public static Task<ImageOcrResult?> OcrImage(this BotContext bot, string url)
+        => bot.ContextCollection.Business.OperationLogic.ImageOcr(url);
+
+    public static Task<ImageOcrResult?> OcrImage(this BotContext bot, ImageEntity entity)
+        => bot.ContextCollection.Business.OperationLogic.ImageOcr(entity);
 }

Lagrange.Core/Internal/Context/Logic/Implementation/OperationLogic.cs

@@ -717,4 +717,28 @@ public async Task<bool> FriendJoinEmojiChain(uint friendUin, uint emojiId, uint
         var results = await Collection.Business.SendEvent(friendJoinEmojiChainEvent);
         return results.Count != 0 && results[0].ResultCode == 0;
     }
+
+    public async Task<string> UploadImage(ImageEntity image)
+    {
+        await Collection.Highway.ManualUploadEntity(image);
+        var msgInfo = image.MsgInfo;
+        if (msgInfo is null) throw new Exception();
+        var downloadEvent = ImageDownloadEvent.Create(Collection.Keystore.Uid ?? "", msgInfo);
+        var result = await Collection.Business.SendEvent(downloadEvent);
+        var ret = (ImageDownloadEvent)result[0];
+        return ret.ImageUrl;
+    }
+
+    public async Task<ImageOcrResult?> ImageOcr(string imageUrl)
+    {
+        var imageOcrEvent = ImageOcrEvent.Create(imageUrl);
+        var results = await Collection.Business.SendEvent(imageOcrEvent);
+        return results.Count != 0 ? ((ImageOcrEvent)results[0]).ImageOcrResult : null;
+    }
+
+    public async Task<ImageOcrResult?> ImageOcr(ImageEntity image)
+    {
+        var imageUrl = await UploadImage(image);
+        return await ImageOcr(imageUrl);
+    }
 }

Lagrange.Core/Internal/Event/Action/ImageOcrEvent.cs

@@ -0,0 +1,25 @@
+using Lagrange.Core.Common.Entity;
+
+namespace Lagrange.Core.Internal.Event.Action;
+
+internal class ImageOcrEvent : ProtocolEvent
+{
+    public string Url { get; } = string.Empty;
+
+    public ImageOcrResult ImageOcrResult { get; }
+
+    private ImageOcrEvent(string url) : base(true) 
+    {
+        Url = url;
+        ImageOcrResult = new ImageOcrResult(new List<TextDetection>(), "");
+    }
+
+    private ImageOcrEvent(int resultCode, ImageOcrResult result) : base(resultCode)
+    {
+        ImageOcrResult = result;
+    }
+
+    public static ImageOcrEvent Create(string url) => new(url);
+
+    public static ImageOcrEvent Result(int resultCode, ImageOcrResult result) => new(resultCode, result);
+}

Lagrange.Core/Internal/Packets/Service/Oidb/Request/OidbSvcTrpcTcp.0xE07_0.cs

@@ -0,0 +1,41 @@
+using ProtoBuf;
+
+#pragma warning disable CS8618
+// ReSharper disable InconsistentNaming
+
+namespace Lagrange.Core.Internal.Packets.Service.Oidb.Request;
+
+[ProtoContract]
+[OidbSvcTrpcTcp(0xE07, 0)]
+internal class OidbSvcTrpcTcp0xE07_0
+{
+    [ProtoMember(1)] public uint Version { get; set; } 
+
+    [ProtoMember(2)] public uint Client { get; set; }
+
+    [ProtoMember(3)] public uint Entrance { get; set; }
+
+    [ProtoMember(10)] public OcrReqBody OcrReqBody { get; set; }
+}
+
+[ProtoContract]
+internal class OcrReqBody
+{
+    [ProtoMember(1)] public string ImageUrl { get; set; } 
+
+    [ProtoMember(2)] public uint LanguageType { get; set; }
+
+    [ProtoMember(3)] public uint Scene { get; set; }
+
+    [ProtoMember(10)] public string OriginMd5 { get; set; }
+
+    [ProtoMember(11)] public string AfterCompressMd5 { get; set; }
+
+    [ProtoMember(12)] public string AfterCompressFileSize { get; set; }
+
+    [ProtoMember(13)] public string AfterCompressWeight { get; set; }
+
+    [ProtoMember(14)] public string AfterCompressHeight { get; set; }
+
+    [ProtoMember(15)] public bool IsCut { get; set; }
+}

Lagrange.Core/Internal/Packets/Service/Oidb/Response/OidbSvcTrpcTcp0xE07_0Response.cs

@@ -0,0 +1,72 @@
+using ProtoBuf;
+
+#pragma warning disable CS8618
+// ReSharper disable InconsistentNaming
+
+namespace Lagrange.Core.Internal.Packets.Service.Oidb.Response;
+
+[ProtoContract]
+internal class OidbSvcTrpcTcp0xE07_0_Response
+{
+    [ProtoMember(1)] public int RetCode { get; set; }
+
+    [ProtoMember(2)] public string ErrMsg { get; set; }
+
+    [ProtoMember(3)] public string Wording { get; set; }
+
+    [ProtoMember(10)] public OcrRspBody OcrRspBody { get; set; }
+}
+
+[ProtoContract]
+internal class OcrRspBody
+{
+    [ProtoMember(1)] public List<TextDetection> TextDetections { get; set; }
+
+    [ProtoMember(2)] public string Language { get; set; }
+
+    [ProtoMember(3)] public string RequestId { get; set; }
+
+    [ProtoMember(101)] public List<string> OcrLanguageList { get; set; }
+
+    [ProtoMember(102)] public List<string> DstTranslateLanguageList { get; set; }
+
+    [ProtoMember(103)] public List<Language> LanguageList { get; set; }
+
+    [ProtoMember(111)] public uint AfterCompressWeight { get; set; }
+
+    [ProtoMember(112)] public uint AfterCompressHeight { get; set; }
+}
+
+[ProtoContract]
+internal class TextDetection
+{
+    [ProtoMember(1)] public string DetectedText { get; set; }
+
+    [ProtoMember(2)] public uint Confidence { get; set; }
+
+    [ProtoMember(3)] public Polygon Polygon { get; set; }
+
+    [ProtoMember(4)] public string AdvancedInfo { get; set; }
+}
+
+[ProtoContract]
+internal class Polygon
+{
+    [ProtoMember(1)] public List<Coordinate> Coordinates { get; set; }
+}
+
+[ProtoContract]
+internal class Coordinate
+{
+    [ProtoMember(1)] public int X { get; set; }
+
+    [ProtoMember(2)] public int Y { get; set; }
+}
+
+[ProtoContract]
+internal class Language
+{
+    [ProtoMember(1)] public string LanguageCode { get; set; }
+
+    [ProtoMember(2)] public string LanguageDesc { get; set; }
+}

Lagrange.Core/Internal/Service/Action/ImageOcrService.cs

@@ -0,0 +1,57 @@
+using Lagrange.Core.Common;
+using Lagrange.Core.Common.Entity;
+using Lagrange.Core.Internal.Event;
+using Lagrange.Core.Internal.Event.Action;
+using Lagrange.Core.Internal.Packets.Service.Oidb;
+using Lagrange.Core.Internal.Packets.Service.Oidb.Request;
+using Lagrange.Core.Internal.Packets.Service.Oidb.Response;
+using Lagrange.Core.Utility.Extension;
+using ProtoBuf;
+
+namespace Lagrange.Core.Internal.Service.Action;
+
+[EventSubscribe(typeof(ImageOcrEvent))]
+[Service("OidbSvcTrpcTcp.0xe07_0")]
+internal class ImageOcrService : BaseService<ImageOcrEvent>
+{
+    protected override bool Build(ImageOcrEvent input, BotKeystore keystore, BotAppInfo appInfo,
+        BotDeviceInfo device, out Span<byte> output, out List<Memory<byte>>? extraPackets)
+    {
+        var packet = new OidbSvcTrpcTcpBase<OidbSvcTrpcTcp0xE07_0>(new OidbSvcTrpcTcp0xE07_0
+        {
+            Version = 1,
+            Client = 0,
+            Entrance = 1,
+            OcrReqBody = new OcrReqBody
+            {
+                ImageUrl = input.Url,
+                OriginMd5 = "",
+                AfterCompressMd5 = "",
+                AfterCompressFileSize = "",
+                AfterCompressWeight = "",
+                AfterCompressHeight = "",
+                IsCut = false
+            }
+        });
+        output = packet.Serialize();
+        extraPackets = null;
+        return true;
+    }
+
+    protected override bool Parse(Span<byte> input, BotKeystore keystore, BotAppInfo appInfo, BotDeviceInfo device,
+        out ImageOcrEvent output, out List<ProtocolEvent>? extraEvents)
+    {
+        var packet = Serializer.Deserialize<OidbSvcTrpcTcpBase<OidbSvcTrpcTcp0xE07_0_Response>>(input);
+        var response = new ImageOcrResult(
+            packet?.Body.OcrRspBody.TextDetections.Select(d => new Common.Entity.TextDetection(
+                d.DetectedText,
+                (int)d.Confidence,
+                d.Polygon.Coordinates.Select(c => new Common.Entity.Coordinate(c.X, c.Y)).ToList() 
+            )).ToList() ?? new List<Common.Entity.TextDetection>(), 
+            packet?.Body.OcrRspBody.Language ?? string.Empty 
+        );
+        output = ImageOcrEvent.Result(packet?.Body.RetCode ?? -1, response);
+        extraEvents = null;
+        return true;
+    }
+}

Lagrange.OneBot/Core/Entity/Action/OneBotOcrImage.cs

@@ -0,0 +1,9 @@
+using System.Text.Json.Serialization;
+
+namespace Lagrange.OneBot.Core.Entity.Action;
+
+[Serializable]
+public class OneBotOcrImage
+{
+    [JsonPropertyName("image")] public string Image { get; set; } = string.Empty;
+}

Lagrange.OneBot/Core/Operation/Ability/UploadImageOperation.cs

@@ -1,6 +1,6 @@
 using System.Text.Json.Nodes;
 using Lagrange.Core;
-using Lagrange.Core.Internal.Event.Message;
+using Lagrange.Core.Common.Interface.Api;
 using Lagrange.Core.Message.Entity;
 using Lagrange.OneBot.Core.Entity.Action;
 using Lagrange.OneBot.Message.Entity;
@@ -15,15 +15,8 @@ public async Task<OneBotResult> HandleOperation(BotContext context, JsonNode? pa
         if (payload?["file"]?.ToString() is { } file && CommonResolver.ResolveStream(file) is { } stream)
         {
             var entity = new ImageEntity(stream);
-            await context.ContextCollection.Highway.ManualUploadEntity(entity);
-            var msgInfo = entity.MsgInfo;
-            if (msgInfo is null) throw new Exception();
-
-            var downloadEvent = ImageDownloadEvent.Create(context.ContextCollection.Keystore.Uid ?? "", msgInfo);
-            var result = await context.ContextCollection.Business.SendEvent(downloadEvent);
-            var ret = (ImageDownloadEvent)result[0];
-
-            return new OneBotResult(ret.ImageUrl, 0, "ok");
+            var url = await context.UploadImage(entity);
+            return new OneBotResult(url, 0, "ok");
         }
 
         throw new Exception();

Lagrange.OneBot/Core/Operation/Generic/OcrImageOperation.cs

@@ -0,0 +1,27 @@
+using System.Text.Json;
+using System.Text.Json.Nodes;
+using Lagrange.Core;
+using Lagrange.Core.Message.Entity;
+using Lagrange.Core.Common.Interface.Api;
+using Lagrange.OneBot.Core.Entity.Action;
+using Lagrange.OneBot.Core.Operation.Converters;
+using Lagrange.OneBot.Message.Entity;
+
+namespace Lagrange.OneBot.Core.Operation.Generic;
+
+
+[Operation(".ocr_image")]
+[Operation("ocr_image")]
+public class OcrImageOperation() : IOperation
+{
+    public async Task<OneBotResult> HandleOperation(BotContext context, JsonNode? payload)
+    {
+        if (payload.Deserialize<OneBotOcrImage>(SerializerOptions.DefaultOptions) is { } data && CommonResolver.ResolveStream(data.Image) is { } stream)
+        {
+            var entity = new ImageEntity(stream);
+            var res = await context.OcrImage(entity);
+            return new OneBotResult(res, 0, "ok");
+        }
+        throw new Exception();
+    }
+}